Refusal Evaluation on WildGuard Harmful
84.35Refusal RateLow-Rank Combination
Evaluation Results
| Method | Links | |
|---|---|---|
| Low-Rank CombinationBackbone=DEEPSEEK R1 DISTILL LLAMA, Method=Low-Rank Combination2026.03 | 84.35 | |
| DEEPSEEK R1 DISTILL LLAMABackbone=DEEPSEEK R1 DISTILL LLAMA2026.03 | 77.72 | |
| Categorical SteeringBackbone=REFUSE-LLAMA, Method=Categorical Steering2026.03 | 77.19 | |
| Low-Rank CombinationBackbone=LLAMA 3 8B INSTRUCT, Method=Low-Rank Combination2026.03 | 74.27 | |
| Low-Rank CombinationBackbone=REFUSE-LLAMA, Method=Low-Rank Combination2026.03 | 73.87 | |
| LLAMA 3 8B INSTRUCTBackbone=LLAMA 3 8B INSTRUCT2026.03 | 73.74 | |
| REFUSE-LLAMABackbone=REFUSE-LLAMA2026.03 | 59.02 |